Faster delivery of faster science workflows using faster hardware building blocks

Digital computing platforms, from sensors to supercomputers, are vital for many branches of research and innovation. Increasingly, they are either accelerated by special-purpose hardware - notably in the realm of AI - or act themselves as accelerators to traditional digital research infrastructure (DRI). Such accelerated compute has the potential to drive transformational change in the frontiers of knowledge through enhanced productivity, yet if, and only if, sufficient expertise is available to translate research ambition into technical solutions.

SHAREing’s contributions to the community

SHAREing will

  1. foster an ecosystem in which RTPs and other stakeholders can design, prototype and assess the best accelerated software-hardware combination for their research, i.e. pioneer solutions.
  2. establish structured learning, including upskilling pathways, such that RTPs can acquire all skills to realise accelerated computing projects and to subsequently train others.
  3. showcase how RTPs’ interventions enable significant research innovation if they transform into dev-ops that lead accelerate computing projects.

SHAREing’s Assessment

SHAREing will establish a UK-wide service where research groups and SMEs can submit their code for assessment. Subject to funding, SHAREing will produce reports addressing three key questions:

  • Is this codebase well-suited for accelerated computing?
  • What is the best-case performance gain that can be expected?
  • Which compute platforms would be best-suited for this use case?

To ensure successful delivery, the consortium must gather information on existing performance assessment and benchmarking approaches. These approaches will then be compiled into a bespoke SHAREing methodology, which will then be delivered to the UK community through SHAREing assessments. All knowledge and documentation will be made publicly available so that any compute centre can benefit from this work. In addition to the SHAREing methodology, the project will collect information on accelerator testbeds from across the UK and publish this information on its websites. Compute centres will have the opportunity to register their testbeds and use SHAREing as a platform to share knowledge about their platforms with the wider community, thereby increasing the likelihood of recruiting adopters and testers.

SHAREing’s Training

SHAREing’s training must cover content along two dimensions: technical training and professional skills training. For both areas, the consortium will bring together and curate existing content, avoiding reinventing the wheel wherever possible. However, where we identify gaps, SHAREing will have resources to commission the creation of new content.

This applies particularly to the professional skills space, where training material is limited. SHAREing will need to determine what specific skills are required for faster delivery of faster science workflows using faster hardware building blocks, and identify where upskilling content can help strengthen the position of digital Research Technology Professionals:

SHAREing’s Lobbying

Digital Research Technology Professionals (RTPs), in dedicated RTP posts or covering RTP aspects as part of their academic or industry roles, either cross-cutting or embedded within domain disciplines, irrespective of specialism (e.g. platforms or software), are positioned at the intersection of research and digital technologies and are critical to this endeavour. In the future, they will lead on research software development, system design, deployment and workflow/user environment management to map academic ambitions onto development and deployment. They will champion the translation into computing solutions. This is becoming increasingly challenging as systems become multifaceted, tailored, and heterogeneous, and as the rate of change is itself accelerating.

RTPs must be able to draw upon deep technical knowledge, horizon scan for new developments, navigate research and service delivery models, and develop strong professional skills to advocate and implement changes across diverse research domains and their home institutions. Despite their critical role, the number of RTPs per institution remains low within the UK, with support for professional development heavily reliant upon grassroots activities among RTPs and initiatives by vendors. This is not sustainable in the long-term. SHAREing will therefore establish a shared UK-wide and cross-disciplinary Hub that supports and enables RTPs to accelerate research and innovation through the transition of large-scale compute-, data-intensive and AI workflows into the era of accelerated compute.

Outcome and realisation

SHAREing focuses predominantly on the required skills, learning ecosystems facilitating training, and the sharing of expertise, material and core initiatives, but also will contribute towards an ecosystem to apply the acquired skills to enable the use of accelerated compute within the UK’s research and innovation landscape. Inclusivity, accessibility and a science-centred ethos will be core to SHAREing, encouraging open-mindedness and critical thinking, respecting that there is often no single `right way’ when operating at the forefront of technology and research. It will deliver a talented workforce with core competencies for accelerated research computing environments. A close collaboration with national, European and international initiatives will strengthen its impact and branding and establish the UK as showcase of the benefits of using a holistic approach to creating a national DRI, where RTPs partner with academics to dev-op the most challenging simulation-, throughput- or AI-workloads.